machine learning - Sizes of training set and test set -

What happens when testing sets are set to small for a particular case when using those types of data sets What are the guessable approaches? Can someone give me some introduction to manage those conditions and how each learning algorithm will work.

What's your concern? What's wrong with you hope arises from that? Too many verification? Overvalidation?

The test set is not included in training the classifier.

It only allows you to predict the quality of your classifier Future data.

So if you have a big test, then you will probably get a better prediction of the quality of your classifier, all this is something.